A camshaft symmetry detection method and system based on three-dimensional laser

Abstract

The application discloses a camshaft symmetry degree detection method and system based on three-dimensional laser, and belongs to the technical field of three-dimensional laser detection. The method obtains three-dimensional laser scanning point cloud data of a camshaft to be detected and encodes according to scanning line order, selects two end supporting journal cylindrical surface point cloud to fit a rotating axis, and transforms the point cloud to a workpiece coordinate system; equal-interval cross sections are divided along an axial direction, density clustering and polar angle sorting are performed on cross section point sets, and a convex hull and concave hull composite algorithm is adopted to extract cross section profiles; the profiles are converted into polar radius sequences and are spline smoothed, and cam peach tip, base circle and buffer segment key points are extracted; a symmetry degree evaluation reference surface is established according to the key points, and a range of distance deviations of all surface points to the reference surface is used as camshaft symmetry degree. The scheme can retain complex cam profile characteristics and utilize all surface points to evaluate symmetry degree, and the detection result is closer to a real geometric deviation.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional laser inspection technology, specifically to a method and system for detecting the symmetry of a camshaft based on three-dimensional laser. Background Technology

[0002] Existing technologies for camshaft symmetry detection are mainly divided into two categories: contact and non-contact. Contact coordinate measuring machines (CMMs) and roundness testers sample point by point along the camshaft surface using a probe, achieving high single-point accuracy. However, they are slow, lack sufficient coverage of the entire camshaft surface, and the measuring force may cause slight deformation of slender shafts. Non-contact laser triangulation and structured light scanning can obtain high-density point clouds, but these point clouds contain outliers caused by surface oil, roughness, and reflections. When extracting the cross-sectional profile, directly using least-squares circle fitting, fixed-threshold filtering, or convex hull operations can easily simplify or mistakenly delete the convex and concave features at the transition between the cam buffer section and the base circle. The cam cross-section is not a single-radius circle; the curvature at the tip is large, the radial change in the buffer section is gentle, and there are also local non-convex regions. Convex hull alone can only obtain the outer boundary and cannot represent the concave part; concave hull alone is sensitive to noise, and the boundary point chain is prone to breakage. In terms of symmetry evaluation, conventional methods often use the center of the cam tip, the center of the base circle, or a fitted center of a certain cross section to construct a benchmark, and then indirectly characterize symmetry through local runout or roundness indices. This approach does not fully utilize the surface points of all cross sections, and misjudgment of individual feature points can significantly affect the evaluation results. Three-dimensional laser point cloud data is large in volume and has many cross sections, requiring the preservation of complex cam cross-sectional contours under noisy conditions, and necessitating the establishment of a symmetry evaluation benchmark that can cover the entire surface based on the main geometric sections of the cam. Summary of the Invention

[0003] To address the shortcomings of existing camshaft symmetry detection methods, such as insufficient handling of non-convex boundaries and outliers in cross-sectional contour extraction and difficulty in reflecting overall surface deviations in symmetry evaluation benchmarks, a 3D laser-based camshaft symmetry detection method and system are proposed. By acquiring a 3D laser-scanned point cloud and fitting a rotation axis, axial cross-sections are divided in the workpiece coordinate system. Density clustering, polar angle sorting, and composite contour extraction of convex and concave hulls are performed on the cross-sectional point set. Key points of the cam tip, base circle, and buffer section are then extracted, and a symmetry evaluation benchmark surface is established. The camshaft symmetry is characterized by the range of distance deviations from all surface points to the benchmark surface, achieving non-contact, full-surface evaluation of camshaft symmetry.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a camshaft symmetry detection method based on three-dimensional laser, comprising: acquiring three-dimensional laser scanning point cloud data of the camshaft to be tested; encoding the three-dimensional laser scanning point cloud data according to the scanning line sequence to obtain an initial point cloud sequence; selecting cylindrical surface point clouds of the journals supporting both ends from the initial point cloud sequence, fitting the rotation axis, and transforming all point clouds to a workpiece coordinate system with the rotation axis as the coordinate axis; setting multiple equally spaced cross sections along the axial direction in the workpiece coordinate system, calculating the cross section index to which each point belongs, and obtaining a cross section point set; performing density clustering on the cross section point set, removing outliers, and then sorting the remaining points according to polar angle to generate a boundary point chain; extracting the cross section contour using a convex hull and concave hull composite algorithm on the boundary point chain; converting the cross section contour into a polar diameter sequence, performing spline smoothing on the polar diameter sequence, and extracting key points of the cam tip, base circle, and buffer section; establishing a symmetry evaluation reference surface based on the key points, calculating the range of the distance deviations of all surface points to the reference surface, and obtaining the camshaft symmetry.

[0005] The above technical solution uses three-dimensional laser scanning point cloud as the data source. By automatically fitting the rotation axis and constructing the workpiece coordinate system, it provides a unified benchmark for subsequent symmetry evaluation, reduces manual alignment errors, and improves the consistency and accuracy of the test results.

[0006] In a preferred embodiment, the process of acquiring 3D laser scanning point cloud data is as follows: A 3D laser scanner is controlled to stop at multiple preset angles and emit line lasers to acquire laser stripe images on the camshaft surface; grayscale centroids are extracted column by column from the laser stripe images, and sub-pixel stripe center coordinates are calculated; the sub-pixel stripe center coordinates are timestamped with the scanner rotation angle and axial displacement to obtain 3D point coordinates; the 3D point coordinates are sorted in ascending order by scanner rotation angle and ascending order by axial displacement to generate the initial point cloud sequence. This process can obtain point clouds with clear line sequences and uniform density, providing reliable data for cross-sectional point set division and contour extraction.

[0007] In a preferred embodiment, the process of fitting the rotation axis is as follows: The cylindrical point cloud supporting the journal at both ends is used as the journal point cloud; principal component analysis is performed on the journal point cloud, and the eigenvector corresponding to the largest eigenvalue is taken as the initial axis; multiple parallel planes are constructed with the direction of the initial axis as the normal; the journal point cloud is projected onto the parallel planes, and circle fitting is performed on the projected points to obtain the fitting circle centers of each plane; connecting the fitting circle centers generates a spatial straight line, which is used as the updated axis; the projection and circle fitting are repeated with the updated axis as the initial axis until the angle between two adjacent updated axes is less than a preset threshold, thus obtaining the rotation axis. Through projection circle fitting and iterative optimization, the influence of local journal shape errors and point cloud noise on the rotation axis can be reduced.

[0008] In a preferred embodiment, the process of generating the cross-sectional point set is as follows: the axial coordinates of all point clouds that have been transformed to the workpiece coordinate system are divided into intervals according to a preset spacing to obtain multiple axial intervals; each point is assigned to the axial interval in which its axial coordinates are located; within each axial interval, the point cloud is divided into multiple sector blocks according to the polar angle, and the number of points and the average radial distance in each sector block are calculated; the number of points and the average radial distance are written into the data table corresponding to the cross-sectional index to generate the cross-sectional point set. This method makes the cross-sectional point set more evenly distributed, which is convenient for subsequent clustering and contour extraction.

[0009] In a preferred embodiment, the process of generating the boundary point chain is as follows: for each point in the cross-sectional point set, count the number of neighboring points within a preset radius centered on that point; mark points with a neighboring point count greater than a first threshold as core points; group core points with a distance less than a second threshold into the same cluster; remove points in clusters with a point count less than a third threshold as outliers; sort the remaining points by polar angle from smallest to largest, and connect adjacent points to generate the boundary point chain. Removing outliers through density clustering effectively suppresses the interference of scanning burrs and stray points on contour extraction.

[0010] In a preferred embodiment, the process of extracting the cross-sectional profile is as follows: convex hull calculation is performed on all points in the boundary point chain to obtain an outer convex hull; the boundary points inside the outer convex hull are marked as concave points, and the concave points are inserted into the corresponding edges of the outer convex hull to obtain a concave hull; the length of each edge of the concave hull is calculated, and the point pairs corresponding to the edges with lengths greater than a fourth threshold are deleted; the deleted points are reordered according to their polar angles, and cubic spline interpolation is used to generate a closed cross-sectional profile. The combined convex and concave hull processing can take into account both the convex profile and the concave details of the cam cross-section, generating a continuous and closed profile line.

[0011] In a preferred embodiment, the process of extracting key points is as follows: taking the center of the corresponding section on the rotation axis as the pole, each point on the cross-sectional contour is converted into a polar radius and polar angle; the polar radius is resampled at polar angle intervals to form a polar radius sequence; the polar radius sequence is smoothed using cubic spline smoothing; based on the local variation characteristics of the smoothed polar radius sequence, the local maxima and local minima of the polar radius are extracted, the point corresponding to the local maxima is taken as the cam tip, the segment with a continuous radial variation rate less than a fifth threshold is taken as the base circle segment, and the transition point between the cam tip and the base circle segment is taken as the key point of the buffer segment. This process can stably identify the key geometric segments of the cam from the smoothed contour, reducing misjudgments caused by local contour fluctuations.

[0012] In a preferred embodiment, the process of establishing the symmetry evaluation reference surface is as follows: key points are projected onto a plane perpendicular to the rotation axis to obtain a set of projected points; principal component analysis is performed on the projected point set, and the direction of the second principal component is taken as the symmetry direction; a symmetry evaluation reference surface is constructed using the rotation axis and the symmetry direction; the distance deviations of all surface points to the symmetry evaluation reference surface are calculated; and the range of the distance deviations is taken as the symmetry of the camshaft. Using the range of the distance deviations of all surface points to the reference surface for evaluation can reflect the symmetry of the overall camshaft profile relative to the reference surface, resulting in a more objective outcome.

[0013] This invention also provides a camshaft symmetry detection system based on three-dimensional laser, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the aforementioned camshaft symmetry detection method based on three-dimensional laser. This system, through three-dimensional laser non-contact scanning, automatic point cloud processing, and symmetry evaluation, features high detection speed, good repeatability, and wide applicability.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: In the cross-sectional contour extraction, density clustering is first performed on the cross-sectional point set to identify core points based on the number of neighborhood points. Clusters with too few points are removed as outliers. Then, the retained points are sorted by polar angle to generate a boundary point chain. This process preserves the main contour of the cam cross-section while suppressing discrete points caused by scanning stripe noise, reflective flying points, and local occlusion, ensuring that subsequent contour generation is not affected by a few outliers. A composite convex hull and concave hull algorithm is used for the boundary point chain. First, the convex hull is calculated to obtain the outer convex hull. Then, the boundary points inside the outer convex hull are inserted as concave points into the corresponding edges to form the concave hull. Excessively long concave hull edges are deleted, and the points are reordered by polar angle and cubic spline interpolation is used to generate a closed cross-sectional contour. Compared with conventional single convex hull or simple concave hull methods, this composite algorithm does not mistakenly fill the smooth concave area and non-convex transition area between the base circle and the buffer section as a convex region. It also avoids over-fitting local noise. The cross-sectional contour retains the boundaries of the tip, base circle, and buffer section more completely, and the polar diameter sequence better reflects the true geometry of the cam cross-section. In symmetry evaluation, the smoothed extreme diameter sequence is used to extract key points of the cam tip, base circle, and buffer section. The zero-crossing points of the first derivative provide local maxima at the tip and local minima at the base circle. Sections with small continuous radial variation rates are identified as base circle sections, and transition points are identified as key points of the buffer section. These key points are projected onto a plane perpendicular to the rotation axis, and principal component analysis is used to obtain the symmetry direction. A symmetry evaluation reference plane is then constructed using the rotation axis and the symmetry direction. This reference plane does not depend solely on the single local position of the tip or base circle but rather on the main geometric sections of the cam profile. The range of distance deviations from all surface points to this reference plane is used as the camshaft symmetry, directly reflecting the maximum dispersion of the entire camshaft surface relative to the reference plane. Compared to evaluation methods that use only a few feature points or a single cross-section center to construct the reference, this method more fully incorporates the profile deviations of each cross-section, reducing the impact of local machining defects or mis-extracted individual feature points on the overall evaluation results. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a flowchart of a camshaft symmetry detection method based on three-dimensional laser. Figure 2 This is a flowchart of the method for generating 3D point clouds of a camshaft; Figure 3 This is a schematic diagram of Gaussian curve fitting of the grayscale intercept of laser stripes; Figure 4 This is a schematic diagram of the point cloud projection and multi-section circle fitting of the journal; Figure 5 This is a diagram showing the distribution of distance deviations from points on the camshaft surface to the symmetry evaluation reference surface. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] See Figure 1 This invention provides a camshaft symmetry detection method based on three-dimensional laser, comprising: acquiring three-dimensional laser scanning point cloud data of the camshaft to be tested; encoding the three-dimensional laser scanning point cloud data according to the scanning line sequence to obtain an initial point cloud sequence; selecting cylindrical surface point clouds of the journals supporting both ends from the initial point cloud sequence, fitting the rotation axis, and transforming all point clouds to a workpiece coordinate system with the rotation axis as the coordinate axis; setting multiple equally spaced cross sections along the axial direction in the workpiece coordinate system, calculating the cross section index to which each point belongs, and obtaining a cross section point set; performing density clustering on the cross section point set, removing outliers, and sorting the retained points by polar angle to generate a boundary point chain; extracting the cross section contour using a convex hull and concave hull composite algorithm on the boundary point chain; converting the cross section contour into a polar diameter sequence, performing spline smoothing on the polar diameter sequence, and extracting key points of the cam tip, base circle, and buffer section; establishing a symmetry evaluation reference surface based on the key points, calculating the range of the distance deviations of all surface points from the reference surface, and obtaining the camshaft symmetry.

[0019] In specific implementation, please refer to Figure 2 A 3D laser scanner emits line lasers at multiple preset angles to acquire laser stripe images of the camshaft surface. The 3D laser scanner includes a rotary table, a line laser emitter, an industrial camera, and an axial displacement mechanism. The camshaft under test is clamped on the rotary table, which rotates the camshaft to multiple preset angles and holds it there. At each preset angle, the line laser emitter projects a laser plane onto the camshaft surface, and the industrial camera captures the modulated laser stripe images of the camshaft surface. At the same preset angle, the axial displacement mechanism moves the line laser emitter and the industrial camera along the camshaft axis, acquiring laser stripe images at multiple axial positions.

[0020] The gray-level centroids are extracted column by column from the laser stripe image, and the sub-pixel stripe center coordinates are calculated. For each column of the laser stripe image, the row of pixels with the largest gray value is found and recorded as the initial center row. In specific implementation, all pixel gray values ​​in the current column of the laser stripe image are traversed, and the coordinates of the row of pixels with the largest gray value are determined by comparison, and these coordinates are recorded as the initial center row.

[0021] A preset number of neighboring rows are selected above and below the initial center row to form the grayscale cutoff line of the current column. The preset number is 7, meaning 7 rows are selected above and 7 rows are selected below the initial center row, totaling 15 rows to form the grayscale cutoff line. The reason for setting the preset number to 7 is that the energy of the laser stripe cross section is concentrated, and 15 rows of grayscale cutoff lines can cover the main distribution area of ​​the laser stripe energy, while avoiding the introduction of too many background pixels.

[0022] Gaussian curve fitting is performed on the grayscale intercept. The Gaussian curve fitting model is as follows: in, Represents the uplink coordinates of the grayscale cutoff line. The fitted gray value at that location; This represents the difference between the peak gray level of the laser stripe cross section and the background gray level. This indicates the peak position of the Gaussian curve, i.e., the center row of the subpixel stripes in the current column; This parameter represents the width of the laser stripe cross section and describes the range of grayscale distribution of the laser stripe. Indicates the background grayscale; This represents the pixel row coordinates within the grayscale cutoff line. During the fitting process, the initial center row is used as... The initial value is the difference between the maximum and minimum gray values ​​in the grayscale cross-section. The initial value is 1.5. The initial value is the minimum gray value in the grayscale cutoff line. The initial value is obtained by iteratively solving the Levenberg-Marquardt algorithm. , , , Four parameters. The solution obtained... As the center row of the subpixel stripes in the current column.

[0023] Based on the line laser scanner calibration parameters, the center row of subpixel stripes and the current column number are converted into three-dimensional point coordinates within the laser plane. In practice, the line laser scanner calibration parameters include the plane equation of the laser plane in the camera coordinate system, the camera intrinsic parameter matrix, distortion coefficients, rotation matrix, and translation vector. These parameters are obtained through planar target calibration and laser plane calibration. For the image coordinates formed by the current column number and the center row of subpixel stripes, the image coordinates are first converted to normalized coordinates using the camera intrinsic parameter matrix and distortion coefficients. Then, the laser plane equation and the ray equation from the camera optical center to the normalized coordinates are solved simultaneously, and the intersection point is obtained to obtain the three-dimensional point coordinates within the laser plane.

[0024] The subpixel stripe center coordinates are timestamped with the scanner's rotation angle and axial displacement to obtain the 3D point coordinates. In practice, when the industrial camera acquires each frame of laser stripe image, it simultaneously records the readings of the rotary table angle encoder and the grating ruler of the axial displacement mechanism, ensuring timestamping alignment through hardware trigger signals. For each subpixel stripe center coordinate, the corresponding scanner rotation angle and axial displacement are matched according to the laser stripe image frame index where the subpixel stripe center coordinates are located. The 3D point coordinates in the laser plane are transformed to the scanner's reference coordinate system using a rotation matrix and translation vector. Then, a rotation and translation transformation is performed according to the corresponding scanner rotation angle and axial displacement to obtain the 3D point coordinates in the scanner's reference coordinate system.

[0025] The 3D point coordinates are sorted in ascending order by both scanner rotation angle and axial displacement to generate an initial point cloud sequence. In practice, a sorting algorithm is used to double-sort all 3D point coordinates based on both scanner rotation angle and axial displacement. First, the scanner rotation angle is used as the primary sorting key for ascending order. Then, if the scanner rotation angles are the same, the axial displacement is used as the secondary sorting key for ascending order. After sorting, an initial point cloud sequence organized according to the scan line order is obtained.

[0026] See Figure 3 In the figure, the horizontal axis represents the pixel row coordinates (in pixels) of the laser stripe image, and the vertical axis represents the gray value of the corresponding pixel row. The blue scatter points represent the gray-level cross-section data points extracted column by column from the laser stripe image, and their distribution exhibits a typical gray-level peak shape of laser stripes. The red curve is the Gaussian curve fitting result of the blue gray-level data points. The fitted curve has a smooth shape and clearly reflects the gray-level peaks of the laser stripes, verifying the good fitting ability of the Gaussian curve model for the gray-level distribution of the laser stripe cross-section.

[0027] The green dashed line in the figure represents the initial center row position, which is the row of pixels with the highest grayscale value determined by traversing the grayscale values ​​of the current column, and serves as the initial value for the Gaussian curve peak parameter β. As can be seen from the figure, the initial center row is located to the left of the grayscale peak, with a certain offset. The pink dashed line represents the sub-pixel stripe center row position calculated through Gaussian curve fitting. This position is located to the right of the initial green center row, close to the theoretical center of the grayscale peak, reflecting the sub-pixel level stripe center extraction accuracy.

[0028] The overall grayscale peak is located in the range of approximately 7 to 9 pixel row coordinates, with a peak grayscale value of approximately 200 or higher. By fitting a red Gaussian curve, the peak position (sub-pixel center) can be accurately extracted, providing basic data for subsequently converting the stripe center into three-dimensional point coordinates within the laser plane.

[0029] The cylindrical point cloud of the journals supporting both ends is used as the journal point cloud. In specific implementation, the cylindrical point cloud of the journals supporting both ends is separated from the initial point cloud sequence according to the axial coordinate range and radial distance range. The separated point clouds are merged into the journal point cloud, which contains the surface points of the journals supporting both ends of the camshaft under test. Principal component analysis is performed on the journal point cloud, and the eigenvector corresponding to the largest eigenvalue is taken as the initial axis. In specific implementation, a three-dimensional coordinate covariance matrix of the journal point cloud is constructed, and the eigenvalues ​​and eigenvectors of the covariance matrix are solved. The eigenvalues ​​are arranged in descending order, and the eigenvector corresponding to the largest eigenvalue is taken as the initial axis. Multiple parallel planes are constructed with the direction of the initial axis as the normal. The journal point cloud is projected onto the parallel planes, and the projected points are fitted with circles to obtain the fitted circle centers of each plane. In specific implementation, the parallel planes are arranged at equal intervals along the initial axis direction within the axial distribution range of the journal point cloud, and the number of parallel planes is set to 20. The number of parallel planes is set to 20 because: 20 parallel planes can form a sufficient number of cross-sections within the axial length of the journal point cloud; the number of projection points on each cross-section is sufficient to meet the stability requirements of circle fitting; and it avoids the projection direction error caused by too small a distance between the fitting circle centers due to too many parallel planes. For each parallel plane, the journal point cloud is projected onto the parallel plane along the initial axial direction to obtain a set of projection points. Least squares circle fitting is used on the set of projection points on each parallel plane, and the objective function for fitting the circle is: in, Indicates the first The x-coordinate of the center of the fitted circle on a parallel plane; Indicates the first The ordinate of the center of the fitted circle on a parallel plane; Indicates the first The radius of the fitted circle on a parallel plane; Indicates the first On the nth parallel plane The x-coordinates of the projection points; Indicates the first On the nth parallel plane The ordinates of the projection points; Indicates the first Number of projection points on a parallel plane; Indicates the number of the parallel plane; Indicates the first The index of the projection points on the parallel plane. Minimize When using the Gauss-Newton iterative method to solve... , , The initial values ​​are taken as the mean coordinates of the projected point set and the average distance from the projected point set to the mean coordinate point. The obtained solution... , As the first The coordinates of the fitted circle centers on all parallel planes are used. Connecting the fitted circle centers generates a spatial straight line, which is then used as the update axis. In practice, the coordinates of the fitted circle centers on all parallel planes are used to form a 3D point set. A 3D straight line is fitted using least squares on this point set. The 3D straight line fitting uses singular value decomposition (SVD). The right singular vector corresponding to the largest singular value of the 3D point set's coordinate covariance matrix is ​​taken as the spatial straight line direction vector. The mean of the 3D point set coordinates is taken as the point through which the spatial straight line passes. The spatial straight line is determined by the spatial straight line direction vector and the point through which it passes, and this spatial straight line is used as the update axis. Projection and circle fitting are repeated using the update axis as the initial axis until the angle between two adjacent update axes is less than a preset threshold, thus obtaining the rotation axis. In practice, after each iteration, the angle between the direction vectors of two adjacent update axes is calculated. The angle is calculated by normalizing the direction vectors of two adjacent update axes, calculating the dot product of the normalized direction vectors, and taking the inverse cosine of the dot product to obtain the angle. The preset threshold is set to 0.001 degrees. The preset threshold of 0.001 degrees is based on the following: when the angle between two adjacent updated axes is less than 0.001 degrees, the offset of the journal point cloud projection position caused by the change in axis direction is less than 0.1% of the midpoint spacing of the journal point cloud. Continued iteration improves the rotation axis direction by less than the numerical range of the journal point cloud midpoint spacing. Therefore, 0.001 degrees is used as the criterion for iteration convergence. When the angle between two adjacent updated axes is less than 0.001 degrees, the most recently updated axis is used as the rotation axis.

[0030] See Figure 4In the figure, the horizontal axis represents the x-coordinate u of the projected point (in millimeters), and the vertical axis represents the y-coordinate v of the projected point (in millimeters). The coordinate system is a planar coordinate system perpendicular to the rotation axis. Light blue scattered dots represent the point set of the journal point cloud after projection onto multiple parallel planes normal to the rotation axis direction. Red solid dots represent the center coordinates of the corresponding parallel planes obtained through least-squares circle fitting. Black solid arcs represent the fitted circles. The figure shows the circle fitting results for multiple parallel planes. The centers of the fitted circles converge to form a spatial curve, which is used for further three-dimensional straight-line fitting to update the axis direction. The distribution of the fitted circle centers and the continuity of the fitted arcs in the figure demonstrate that the circle fitting process accurately reflects the cylindrical surface characteristics of the journal point cloud, providing a reliable basis for the accurate determination of the subsequent rotation axis. The circles in the figure are linearly arranged along a certain direction, conforming to the initial axis direction obtained from principal component analysis, reflecting the process of the recursive iterative projection and circle fitting steps gradually converging to the rotation axis as described in the above embodiment. The stability of the overall fitted circle radius and center position indicates that the method can effectively eliminate noise points and outliers, ensuring the accuracy of axis fitting.

[0031] In practical implementation, the axial coordinates of all point clouds transformed into the workpiece coordinate system are divided into intervals according to a preset spacing, resulting in multiple axial intervals. The preset spacing is set to 0.5 mm. The basis for setting the preset spacing to 0.5 mm is that the axial point spacing of the 3D laser scanning point cloud on the camshaft surface is approximately 0.2 mm. A 0.5 mm axial interval can contain point clouds of two to three scanning lines, ensuring that the number of points in each axial interval meets the requirements for cross-sectional contour extraction. At the same time, the width of the axial interval is less than the minimum wavelength of the axial change of the camshaft cross-sectional contour, avoiding the mixing of adjacent cross-sectional contour information. The axial interval division starts from the minimum axial coordinate of all point clouds in the workpiece coordinate system and is divided sequentially in steps of 0.5 mm until the maximum axial coordinate of all point clouds is covered. Each axial interval is determined by the left boundary axial coordinate and the right boundary axial coordinate. Each point is assigned to the axial interval in which its axial coordinate is located. In practical implementation, for each 3D point in the workpiece coordinate system, the axial coordinate value of the 3D point is obtained, and the 3D point is assigned to the axial interval in which the axial coordinate value satisfies that the left boundary axial coordinate is less than or equal to the right boundary axial coordinate. Within each axial interval, the point cloud is divided into multiple sector blocks according to the polar angle. The number of points and the average radial distance in each sector block are calculated. In practice, the polar angle refers to the azimuth angle of a 3D point relative to the center of the cross-section in a plane perpendicular to the rotation axis. The center of the cross-section is the intersection of the rotation axis and the central plane of the axial interval. For each point within each axial interval, the polar angle is calculated, ranging from 0 radians to 2π radians. The polar angle range is divided into 360 sector blocks, with each sector block corresponding to a polar angle interval of [missing information]. Radius. The number of sector blocks is set to 360 because: 360 sector blocks correspond to one sector block per degree; the polar angle resolution meets the requirements for representing the polar diameter variation of the cam cross-section profile; and each sector block has enough points to participate in the radial distance mean statistics. For each sector block, the number of all points within the sector block is counted as the point count. The radial distance from all points within the sector block to the center of the cross-section is calculated and its arithmetic mean is obtained. The point count and the radial distance mean are written into the data table corresponding to the cross-section index to generate the cross-section point set. In specific implementation, the data table uses the cross-section index and sector block number as the primary key. The cross-section index is the sequential number of the axial interval, and the sector block number is the sequential number of the sector block in the polar angle direction. Each row of the data table records the point count and the radial distance mean corresponding to a sector block. All points within each sector block are retained in the cross-section point set, and the point count and radial distance mean corresponding to the sector block are recorded in the data table. For each point in the cross-section point set, the number of neighboring points within a preset radius centered on the current point is counted. In practical implementation, the preset radius is set to 0.5 mm. The basis for this setting is that the average point spacing of the 3D laser scan point cloud on the camshaft surface is approximately 0.15 mm, and a 0.5 mm radius can cover about 35 points, enabling stable local density estimation in density clustering. For each point in the cross-sectional point set, the Euclidean distance between the current point and other points in the set is calculated. The Euclidean distance calculation formula is: in, Represents the first point in the cross-section point set. The point and the first Euclidean distance between points; Represents the first point in the cross-section point set. The x-coordinate of a point in the workpiece coordinate system; Represents the first point in the cross-section point set. The ordinate of a point in the workpiece coordinate system; Represents the first point in the cross-section point set. The axial coordinates of a point in the workpiece coordinate system; Represents the first point in the cross-section point set. The x-coordinate of a point in the workpiece coordinate system; Represents the first point in the cross-section point set. The ordinate of a point in the workpiece coordinate system; Represents the first point in the cross-section point set. The axial coordinates of each point in the workpiece coordinate system are calculated. The number of other points with an Euclidean distance less than 0.5 mm is counted to obtain the number of neighboring points of the current point. Points with a neighbor number greater than a first threshold are marked as core points. In this implementation, the first threshold is set to 8. The basis for setting the first threshold to 8 is that within a 0.5 mm radius, effective cross-sectional profile points typically have more than 8 neighboring points, while isolated outliers have fewer than 8 neighboring points within a 0.5 mm radius. A first threshold of 8 can distinguish between core points and outliers. For each point in the cross-sectional point set, the corresponding number of neighboring points is obtained, and points with a neighbor number greater than 8 are marked as core points. Points with a distance less than a second threshold between core points are grouped into the same cluster. In this implementation, the second threshold is set to 0.8 mm. The basis for setting the second threshold to 0.8 mm is that 0.8 mm is greater than the average distance between adjacent core points and less than the minimum interval between different feature regions of the cam cross-sectional profile. This allows core points belonging to the same continuous region to be grouped into the same cluster, avoiding the incorrect merging of core points from different regions. Traverse all core points. For each un-clustered core point, create a new cluster, add the current core point to the new cluster, and find other core points whose Euclidean distance to any core point in the new cluster is less than 0.8 mm. Add these other core points to the new cluster. Repeat this process until no new core points are added, completing one cluster. Repeat the above process until all core points are assigned to their respective clusters. Points in clusters with a number of points less than a third threshold are removed as outliers. In this implementation, the third threshold is set to 10. The reason for setting the third threshold to 10 is that the number of points in the continuous region of the cam cross-section profile is usually greater than 10, while the number of points in clusters formed by isolated points or minor disturbances is less than 10. Setting the third threshold to 10 can remove clusters formed by isolated points, preserving the true cross-section profile region. Count the number of points within each cluster. Remove all points in clusters with less than 10 points as outliers, excluding them from subsequent boundary point chain generation. The remaining points are sorted in ascending order of their polar angles, and adjacent points are connected to form a chain of boundary points. In practice, after removing outliers, the polar angle of each remaining point is recalculated, ranging from 0 radians to 2π radians. The remaining points are then arranged in ascending order of their polar angles using a sorting method, resulting in an ordered point sequence. Adjacent points in the ordered sequence are connected sequentially with straight line segments, and the last point is connected to the first point with a straight line segment, forming a closed chain of boundary points.

[0032] In the specific implementation, convex hull calculation is performed on all points in the boundary point chain to obtain the outer convex hull. The convex hull calculation uses the Andrew monotonic chain algorithm, arranging all points in the boundary point chain in ascending order of their x-coordinates, and if the x-coordinates are the same, in ascending order of their y-coordinates. Then, the lower and upper convex hulls are constructed sequentially, and the lower and upper convex hulls are merged to obtain the outer convex hull. Boundary points inside the outer convex hull are marked as concave points. In the specific implementation, for each point in the boundary point chain, the ray method is used to determine whether the point is inside the outer convex hull. A ray is drawn from the current point in the positive x-coordinate direction, and the number of intersections between the ray and the boundary of the outer convex hull is counted. When the number of intersections is odd, the current point is marked as a concave point. The concave point is inserted into the corresponding edge of the outer convex hull to obtain the concave hull. In the specific implementation, for each edge of the convex hull, find the concave point with the largest vertical distance from the edge among the concave points corresponding to that edge, insert this concave point between the two endpoints of the edge to form two new edges, and then update the convex hull to the boundary after inserting the concave point. Repeat the above insertion process until no concave point is located outside the current boundary, thus obtaining the concave hull. Calculate the length of each edge of the concave hull, and delete the point pairs corresponding to edges with a length greater than a fourth threshold. In the specific implementation, the fourth threshold is set to 2 mm. The basis for setting the fourth threshold to 2 mm is that the circumferential spacing between adjacent points of the cam cross-section profile is within 0.5 mm. Edges in the concave hull with a side length greater than 2 mm indicate that the two points connected by the edge cross the blank area of ​​the profile, which is an incorrect connection. Setting the fourth threshold to 2 mm can delete incorrectly connected edges that cross the blank area while retaining the true profile edges. For each edge in the concave hull, calculate the Euclidean distance between the two endpoints corresponding to the edge to obtain the length of each edge of the concave hull, and delete the two endpoints corresponding to the edge with a side length greater than 2 mm from the point set. Reorder the deleted points according to the polar angle, and use cubic spline interpolation to generate a closed cross-sectional profile. In the specific implementation, taking the center of the corresponding section on the rotation axis as the pole, the polar angles of the remaining points after deletion are recalculated. The remaining points are arranged in ascending order of polar angle, and a 2π radian periodic closure condition is added between the first and last points. Cubic spline interpolation is used to construct piecewise cubic polynomials for the sorted remaining points. Each piecewise cubic polynomial passes between two adjacent remaining points and satisfies the continuity of the first and second derivatives at the connection between adjacent pieces. The periodic continuity condition is satisfied between the first and last pieces, resulting in a closed cross-sectional profile. Taking the center of the corresponding section on the rotation axis as the pole, each point on the cross-sectional profile is converted into polar radius and polar angle. In the specific implementation, for each point on the cross-sectional profile, the Euclidean distance from the point to the center of the corresponding section on the rotation axis is calculated as the polar radius, and the angle between the line connecting the point and the center of the corresponding section on the rotation axis in the plane perpendicular to the rotation axis and the positive direction of the horizontal axis is calculated as the polar angle. The polar radius and polar angle constitute the polar coordinate representation of the point. The polar radius is resampled at polar angle intervals to form a polar radius sequence. In the specific implementation, the polar angle interval is taken as 0.01 radians.The polar angle interval of 0.01 radians is chosen because 0.01 radians corresponds to approximately 0.57 degrees. After resampling, the polar angle resolution can distinguish the minimum polar angle change characteristics of the cam buffer section and the base circle section. Simultaneously, the number of resampling points is within a reasonable range; an excessively large polar angle interval will lose details of the buffer section, while an excessively small interval will introduce interpolation noise. A resampled polar angle sequence is generated within the range of 0 to 2π radians at 0.01 radian intervals. For each resampled polar angle, the two adjacent polar angles on the cross-sectional profile are found, and linear interpolation is used to obtain the polar radius corresponding to the resampled polar angle. All polar radii corresponding to the resampled polar angles are combined into a polar radius sequence. Cubic spline smoothing is applied to the polar radius sequence to obtain a smoothed polar radius sequence. In specific implementation, a cubic spline smoothing function is constructed for the polar radius sequence. The cubic spline smoothing function is a piecewise cubic polynomial between adjacent resampled polar angles, satisfying continuous function values, continuous first derivatives, and continuous second derivatives at the junctions of each segment. Periodic continuity conditions are applied at 0 and 2π radians. The piecewise cubic polynomial expression is: in, Indicates the polar angle of resampling The smoothed polar radius obtained by cubic spline smoothing is in millimeters. Indicates the first The initial resampling polar angle of the cubic spline segment, in radians; Indicates the first The constant term coefficients of the cubic spline, in millimeters; Indicates the first The coefficient of the first term of a cubic spline, in millimeters per radian; Indicates the first The quadratic coefficient of a cubic spline segment, expressed in millimeters per radian square; Indicates the first The coefficients of the cubic term in the cubic spline are expressed in millimeters per cubic radian. Each coefficient is obtained by solving a system of linear equations, and its dimension is determined by the difference in polar angles in the corresponding power terms. The power of the term is determined to ensure that the dimension of each power term is millimeters. When solving for the coefficients of each cubic spline segment, the polar radius values ​​in the polar radius sequence are used as the nodal function values. A system of linear equations is established using the conditions of continuous function values ​​at the nodes, continuous first derivative, continuous second derivative, and continuous first and last periods. Solving the system of linear equations yields the coefficients of each cubic spline segment, and then a smooth polar radius sequence is obtained by calculating using piecewise cubic polynomials. Local maxima and local minima of the polar radius are extracted at the zero-crossing points of the first derivative. In the specific implementation, the first derivative function is obtained by differentiating the piecewise cubic polynomial of the smooth polar radius sequence. The sign change of the first derivative function value is detected between two adjacent resampled polar angles. When the first derivative function value changes from positive to negative within an adjacent interval, a local maximum of the polar radius exists within that interval. The position where the first derivative function equals zero is determined using the bisection method, and the smooth polar radius corresponding to this position is taken as the local maximum of the polar radius. When the first derivative function value changes from negative to positive within an adjacent interval, a local minimum of the polar radius exists within that interval. The position where the first derivative function equals zero is determined using the bisection method, and the smooth polar radius corresponding to this position is taken as the local minimum of the polar radius. The cross-sectional profile point corresponding to the local maximum of the polar radius is taken as the cam tip point. The segment with a continuous radial change rate less than the fifth threshold is taken as the base circle segment. In the specific implementation, the fifth threshold is set to 0.02 mm per radian. The fifth threshold of 0.02 mm per radian is chosen because: the polar diameter of the base circle segment changes smoothly with the polar angle, and the radial change rate between adjacent resampled polar angles is less than 0.02 mm per radian. However, the radial change rate near the buffer section and the cam tip is significantly greater than 0.02 mm per radian. Setting the fifth threshold to 0.02 mm per radian distinguishes the base circle segment from the buffer section and the section near the cam tip. The radial change rate between adjacent resampled polar angles is defined as the ratio of the difference in smoothed polar diameters corresponding to two adjacent resampled polar angles to the resampled polar angle interval of 0.01 radian. When the radial change rate between multiple consecutive adjacent polar angle intervals is less than 0.02 mm per radian, these adjacent polar angle intervals are merged into a single base circle segment. The transition point between the cam tip and the base circle segment is designated as the key point of the buffer section. In specific implementation, on the smooth polar diameter sequence, taking the polar angle position corresponding to each cam tip as the starting point, the two nearest base circle segment boundaries are found along the polar angle increasing direction and the polar angle decreasing direction, respectively. The base circle segment boundary is the polar angle position where the radial change rate changes from less than 0.02 mm per radian to greater than 0.02 mm per radian. The turning point where the absolute value of the radial change rate changes from increasing to decreasing in the smooth polar diameter sequence between the cam tip and the base circle segment boundary is taken as the key point of the buffer section.

[0033] In practical implementation, key points are projected onto a plane perpendicular to the rotation axis, resulting in a projection point set. Key points include the cam tip, key points in the buffer section, and key points selected from the base circle section. The projection plane is a plane passing through the origin of the workpiece coordinate system and whose normal is parallel to the rotation axis. For each key point, its spatial coordinates are projected onto the projection plane to obtain a projection point. The projection point retains only the two coordinate components within the plane perpendicular to the rotation axis, and all projection points constitute the projection point set. Principal component analysis is performed on the projection point set, with the direction of the second principal component taken as the symmetry direction. In practical implementation, a two-dimensional coordinate covariance matrix of the projection point set is constructed, and the eigenvalues ​​and eigenvectors of the two-dimensional coordinate covariance matrix are solved. The formula for calculating the two-dimensional coordinate covariance matrix is: in, The two-dimensional coordinate covariance matrix representing the projected point set; This indicates the total number of projection points in the projection point set; Indicates the sequence number of the projection points in the projection point set; Represents the first point in the projection point set. A two-dimensional coordinate column vector of projection points; This represents a two-dimensional column vector of the mean coordinates of all projected points in the projection point set. Indicates the first A centered two-dimensional coordinate column vector of each projection point; Indicates the first The transpose row vector of the centered 2D coordinate column vector of each projection point. Solve for the 2D coordinate covariance matrix. The eigenvalues ​​and eigenvectors are used to obtain two eigenvalues ​​and two corresponding eigenvectors. The eigenvalues ​​are arranged in descending order of value. The eigenvector corresponding to the largest eigenvalue is the direction of the first principal component, and the eigenvector corresponding to the second eigenvalue is the direction of the second principal component. The direction of the second principal component is used as the symmetry direction. A symmetry evaluation reference surface is constructed using the rotation axis and the symmetry direction. In the specific implementation, the rotation axis is the previously fitted rotation axis, and the symmetry direction is the direction of the second principal component. Any point on the rotation axis is taken as the reference surface passage point. The cross product of the rotation axis direction vector and the symmetry direction vector is used as the reference surface normal vector. The reference surface passes through the rotation axis, and the reference surface normal vector is perpendicular to both the rotation axis direction vector and the symmetry direction vector. The symmetry evaluation reference surface is determined by the reference surface passage point and the reference surface normal vector. The distance deviation from all surface points to the symmetry evaluation reference surface is calculated. In the specific implementation, all surface points refer to all surface point clouds on the camshaft under test in the workpiece coordinate system, excluding the point clouds of the end support journals. For each surface point, the directed distance from the surface point to the symmetry evaluation reference plane is calculated. This directed distance is obtained by projecting the vector formed by the surface point coordinates and the reference plane through the point coordinates onto the normal vector of the reference plane. The absolute value of the projection is the distance deviation from the surface point to the symmetry evaluation reference plane. The range of these distance deviations is used as the camshaft symmetry. In a specific implementation, the maximum and minimum values ​​of the distance deviations from all surface points to the symmetry evaluation reference plane are taken. The difference between the maximum and minimum values ​​is the range of the distance deviations, which is used as the camshaft symmetry. A camshaft symmetry detection system based on three-dimensional laser includes a memory, a processor, and a computer program stored in the memory and running on the processor. In a specific implementation, the memory stores three-dimensional laser scanning point cloud data, cross-sectional point sets, boundary point chains, cross-sectional contours, polar diameter sequences, key point coordinates, rotation axis parameters, symmetry evaluation reference plane parameters, and the computer program. When the processor executes the computer program, it implements the steps of the aforementioned detection method, including acquiring three-dimensional laser scanning point cloud data of the camshaft to be tested and encoding the three-dimensional laser scanning point cloud data to obtain an initial point cloud sequence; selecting the cylindrical surface point cloud of the journals supporting both ends from the initial point cloud sequence to fit the rotation axis and transforming all point clouds to the workpiece coordinate system with the rotation axis as the coordinate axis; setting multiple equally spaced sections along the axial direction in the workpiece coordinate system and calculating the section index to which each point belongs to obtain a section point set; performing density clustering on the section point set to remove outliers and sorting the remaining points by polar angle to generate a boundary point chain; using a convex hull and concave hull composite algorithm to extract the section contour on the boundary point chain; converting the section contour into a polar diameter sequence and performing spline smoothing on the polar diameter sequence to extract key points of the cam tip, base circle, and buffer section; establishing a symmetry evaluation reference surface based on the key points, calculating the range of the distance deviations of all surface points to the symmetry evaluation reference surface, and obtaining the camshaft symmetry.

[0034] See Figure 5In the figure, the horizontal axis represents the serial number of the machined surface point on the camshaft under test, and the vertical axis represents the distance deviation from the corresponding surface point to the symmetry evaluation reference surface constructed according to the above embodiment, in millimeters. The scattered blue dot cloud in the figure represents the specific measured value of the distance deviation of each surface point, showing that the deviation of the surface point from the reference surface changes periodically in an approximately sinusoidal form, with the deviation amplitude mainly distributed between -1.0 mm and 1.0 mm, indicating that there is a certain symmetry error fluctuation on the camshaft surface. The green dashed line represents the minimum distance deviation, approximately -1.05 mm, and the red dashed line represents the maximum distance deviation, approximately 1.1 mm. The two constitute the upper and lower boundaries of the distance deviation range. The orange dashed line is located at 0 mm, serving as the zero reference line for the distance deviation. The distance deviation range is calculated from the difference between the maximum and minimum values, approximately 2.15 mm, reflecting the specific value of the symmetry of the camshaft.

[0035] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for detecting the symmetry of a camshaft based on three-dimensional laser, characterized in that, include: Acquire the three-dimensional laser scanning point cloud data of the camshaft to be tested, and encode the three-dimensional laser scanning point cloud data according to the scan line sequence to obtain the initial point cloud sequence; Select the cylindrical surface point cloud of the journal bearings at both ends from the initial point cloud sequence, fit the rotation axis, and transform all point clouds to the workpiece coordinate system with the rotation axis as the coordinate axis; Multiple equally spaced sections are set along the axial direction in the workpiece coordinate system, and the section index of each point is calculated to obtain the section point set; Density clustering is performed on the set of cross-sectional points to remove outliers, and the remaining points are sorted by polar angle to generate a boundary point chain. The cross-sectional contour is extracted using a combined convex hull and concave hull algorithm for the boundary point chain; The cross-sectional profile is converted into a polar radius sequence, and the polar radius sequence is spline smoothed to extract the key points of the cam tip, base circle, and buffer section. Based on the key points, a symmetry evaluation reference surface is established, and the range of the distance deviations of all surface points from the reference surface is calculated to obtain the camshaft symmetry.

2. The camshaft symmetry detection method based on three-dimensional laser as described in claim 1, characterized in that, The process involves acquiring three-dimensional laser scanning point cloud data of the camshaft under test, encoding the three-dimensional laser scanning point cloud data according to the scan line sequence to obtain an initial point cloud sequence, including: Control the 3D laser scanner to stay at multiple preset angles and emit line lasers to acquire laser stripe images on the camshaft surface; Extract the gray-level centroid column by column from the laser stripe image and calculate the center coordinates of the subpixel stripes; The coordinates of the subpixel stripe center are aligned with the scanner rotation angle and axial displacement using timestamps to obtain three-dimensional point coordinates. The three-dimensional point coordinates are sorted in ascending order by scanner rotation angle and axial displacement to generate the initial point cloud sequence.

3. The camshaft symmetry detection method based on three-dimensional laser as described in claim 1, characterized in that, The step of selecting the cylindrical point cloud with two supporting journals from the initial point cloud sequence and fitting the rotation axis includes: The point cloud of the cylindrical surface supporting the journal at both ends is taken as the journal point cloud. Principal component analysis was performed on the point cloud of the journal shaft, and the eigenvector corresponding to the largest eigenvalue was taken as the initial axis. Multiple parallel planes are constructed with the initial axis direction as the normal. The journal point cloud is projected onto the parallel planes, and the projected points are fitted with circles to obtain the fitted circle centers of each plane. Connect the centers of the fitted circles to generate a spatial straight line, and use the spatial straight line as the update axis. Repeat the projection and circle fitting with the updated axis as the initial axis until the angle between two adjacent updated axes is less than a preset threshold, and obtain the rotation axis.

4. The camshaft symmetry detection method based on three-dimensional laser as described in claim 1, characterized in that, The process involves setting multiple equally spaced cross sections along the axial direction in the workpiece coordinate system, calculating the cross section index for each point, and obtaining a set of cross section points, including: The axial coordinates of all point clouds that have been transformed to the workpiece coordinate system are divided into intervals according to a preset spacing to obtain multiple axial intervals; Each point is assigned to the axial interval containing its axial coordinate. Within each axial interval, the point cloud is divided into multiple sector blocks according to the polar angle, and the average number of points and radial distance in each sector block are calculated. The number of points and the average radial distance are written into the data table corresponding to the cross-section index to generate the cross-section point set.

5. The camshaft symmetry detection method based on three-dimensional laser as described in claim 1, characterized in that, The step of performing density clustering on the cross-sectional point set, removing outliers, and then sorting the remaining points by polar angle to generate a boundary point chain includes: For each point in the set of cross-section points, count the number of neighboring points within a preset radius centered on that point; Points with a neighborhood number greater than the first threshold are marked as core points; Points whose distance between core points is less than the second threshold are grouped into the same cluster; Points in clusters with a number of points less than the third threshold are removed as outliers; Sort the remaining points in ascending order of polar angle, and connect adjacent points to generate the boundary point chain.

6. The camshaft symmetry detection method based on three-dimensional laser as described in claim 1, characterized in that, The step of extracting the cross-sectional contour using a combined convex and concave hull algorithm for the boundary point chain includes: Perform convex hull calculation on all points in the boundary point chain to obtain the outer convex hull; Mark the boundary points inside the convex hull as concave points, and insert the concave points into the corresponding edges of the convex hull to obtain the concave hull; Calculate the length of each side of the concave hull, and delete the point pairs corresponding to the sides that are greater than the fourth threshold; The deleted points are reordered according to their polar angles, and cubic spline interpolation is used to generate a closed cross-sectional profile.

7. The camshaft symmetry detection method based on three-dimensional laser as described in claim 1, characterized in that, The process of converting the cross-sectional profile into a polar radius sequence, performing spline smoothing on the polar radius sequence, and extracting key points of the cam tip, base circle, and buffer segment includes: Using the center of the cross section corresponding to the axis of rotation as the pole, each point on the cross section profile is converted into a polar radius and a polar angle; The polar radii are resampled at polar angle intervals to form a polar radii sequence; The polar radius sequence is smoothed using cubic spline smoothing to obtain a smoothed polar radius sequence; Local maxima and local minima of the extreme radius are extracted at the zero point of the first derivative. The point corresponding to the local maximum is taken as the cam tip. The segment with continuous radial change rate less than the fifth threshold is taken as the base circle segment. The transition point between the cam tip and the base circle segment is taken as the key point of the buffer segment.

8. The camshaft symmetry detection method based on three-dimensional laser as described in claim 7, characterized in that, The process of establishing a symmetry evaluation reference surface based on the key points, calculating the range of distance deviations from all surface points to the reference surface, and obtaining the camshaft symmetry includes: Project the key points onto a plane perpendicular to the axis of rotation to obtain the set of projected points; Principal component analysis is performed on the projection point set, and the direction of the second principal component is taken as the symmetry direction; A symmetry evaluation reference surface is constructed using the rotation axis and the symmetry direction; Calculate the distance deviation from all surface points to the symmetry evaluation reference plane; The range of the distance deviation is taken as the symmetry of the camshaft.

9. The camshaft symmetry detection method based on three-dimensional laser according to claim 2, characterized in that, Extracting the grayscale centroid column by column from the laser stripe image and calculating the sub-pixel stripe center coordinates includes: For each column in the laser stripe image, find the row of pixels with the largest gray value and record it as the initial center row; A preset number of neighboring rows are selected above and below the initial center row to form the grayscale cutoff line of the column; Gaussian curve fitting is performed on the grayscale intercept, and the peak position of the fitted curve is taken as the center row of the subpixel stripes of that column; Based on the calibration parameters of the line laser scanner, the center row and column number of the subpixel stripe are converted into three-dimensional point coordinates in the laser plane.

10. A camshaft symmetry detection system based on three-dimensional laser, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the camshaft symmetry detection method based on three-dimensional laser as described in any one of claims 1 to 9.

Images